vix.ing · top · new · best · stats

Optimisation of the event-based TOF filtered back-projection for online imaging in total-body J-PET

2021/07/24 by R. Y. Shopa, K. Klimaszewski, P. Kopka +27 · 18 citations
Mathematics · Medicine · Physics and Astronomy · #Advanced MRI Techniques and Applications #Algorithm #Artificial intelligence #Computer science #Computer vision #Image (mathematics) #Image quality #Image resolution #Imaging phantom #Iterative reconstruction #Kernel (algebra) #Mathematics #Medical Imaging Techniques and Applications #Monte Carlo method #Nuclear medicine #Optics #Parametric statistics #Physics #Positron emission tomography #Radiation Detection and Scintillator Technologies #Radon transform #Scanner #Statistics #Whole body imaging #physics.comp-ph #physics.ins-det #physics.med-ph

paper · pdf · doi:10.1016/j.media.2021.102199

published in Medical Image Analysis 73, 102199 (Elsevier BV)

openalex publication_date 2021/07/24 · arxiv created 2021/07/27 · arxiv updated 2021/08/31 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05

Abstract

We perform a parametric study of the newly developed time-of-flight (TOF) image reconstruction algorithm, proposed for the real-time imaging in total-body Jagiellonian PET (J-PET) scanners. The asymmetric 3D filtering kernel is applied at each most likely position of electron-positron annihilation, estimated from the emissions of back-to-back γ-photons. The optimisation of its parameters is studied using Monte Carlo simulations of a 1-mm spherical source, NEMA IEC and XCAT phantoms inside the ideal J-PET scanner. The combination of high-pass filters which included the TOF filtered back-projection (FBP), resulted in spatial resolution, 1.5 times higher in the axial direction than for the conventional 3D FBP. For realistic 10-minute scans of NEMA IEC and XCAT, which require a trade-off between the noise and spatial resolution, the need for Gaussian TOF kernel components, coupled with median post-filtering, is demonstrated. The best sets of 3D filter parameters were obtained by the Nelder-Mead minimisation of the mean squared error between the resulting and reference images. The approach allows training the reconstruction algorithm for custom scans, using the IEC phantom, when the temporal resolution is below 50 ps. The image quality parameters, estimated for the best outcomes, were systematically better than for the non-TOF FBP.

Citations